ESL-55: A View Egyptian Sign-Language Corpus and the esl_recognition Toolkit for Lightweight Sign Classification

This record provides a Python-based toolkit, esl_recognition, designed to enable machine learning classification of Egyptian Sign Language (ESL) gestures. It accompanies the public dataset "Egyptian Sign Language" (Mendeley Data, V2, doi:10.17632/39tbt2jd7r.2), which contains over 55 labeled sign videos annotated with 33 pose landmarks and 21 hand landmarks per hand using Mediapipe.

The esl_recognition package offers a modular and reproducible workflow for:

• extracting pose and hand landmarks from .mp4 videos,

• preprocessing features (x, y, z, visibility),

• training models like Random Forests or Logistic Regression,

• evaluating performance via cross-validation,

• visualizing metrics, confusion matrices, SHAP plots, and ROC curves,

• running experiments via command-line or Python API.

The toolkit is lightweight, requiring only classic ML libraries, and includes a CLI interface for reproducible experiments using TOML configuration files.

This is one of the first open-source frameworks tailored for Egyptian Sign Language research, lowering the barrier for academic and applied work in assistive communication and accessibility technology for the deaf and hard-of-hearing community in Egypt.